Language Mechanics (technical)

[Nom]

If I may ask, how is the foreign language text generated? From what I can tell, the mapping for a given language is determinate but arbitrary and non-reversible.

My guess is that each source word is used as a seed for a RNG algorithm, the result of which is then re-mapped into a set of syllables. It appears some form of complexity metric is applied to determine which words are "heard" scrambled and which are heard raw.

Is anyone who actually knows willing to comment?

Thanks

[Atama]

Only a programmer can tell you. One of the Skots engineers. Don't count on it. ;)

All I can say is this... They are consistent, in that one word always means another. Also, sometimes multiple foreign words mean one word in Common.

[Ra'Dorcha]

I think you mistyped that. Isn't it multiple common words equal one foreign word? There are plenty of IC reasons that can be given for this.

idly mused by Ra'

[Dariel]

I think he is refering to the fact that one common word -word as in a length of letters strung together- can become two words in the other language. Both Teanga and Western seem to do that from time to time (ie. only some words become two, others translate as one word), but not Northern, as far as I remember; my limited exposure to Eastern and Eme means I cannot really gauge if it happens in those languages.

Otherwise you are of course right, Ra'.


Have a nice day.

[Atama]

Yes, you're right Ra. I think my explanation was a bit confusing.

[Ra'Dorcha]

Dariel is of course also correct. Teanga, from my experience, is a horrid example of that. I have seen Teanga words which translate from one common word into 3 teanga words.

I also doubt you will get the language coders to give you the inside scoop on how they did it. But good luck with the question.

Ra'

[Nom]

I'm pretty confident I understand the gross algorithm. Perhaps what I'd really like to see is an article talking about the policy decisions used; why this mechanism was chosen, how they made the sounds flow into sentences, the comprehension breakpoints, that sort of thing.

The other odd thing I've noticed is that someone who knows the language poorly and yet speaks it STILL produces the correct mapping. IE: the mapped speech is correct, but others do not understand it (see it unmapped). A more stylish option would be to introduce a random factor into the translation for misspoken words. Thus, the words spoken by novices would truly be gibberish.

[Angron]

The languages are basically having the special 'language' syllables replace English/Common syllables to recreate the new word in the foriegn language. There are also a few twists in some of the languages that allow it to seem more realistic (For example, the word 'time' always translates into another word in Teanga). If you want to learn more about it, you can email me at lordarkanor@crosswinds.net - the overall system is quite ingenious, especially for the three languages that were extensively developed in the game.

- Zack

[ChristopherA]

Originally posted by Angron
...especially for the three languages that were extensively developed in the game.

Daris says "The quick brown fox jumped over the lazy dog."

Daris says in Eastern "Or'un iswainel nulnul eridoningol winianer ad or'un is oninis'un."

Daris says in Western "Mimum um rumimum um imim flimgumlum umlumum flim mimum um rum gumrumum um."

Daris says in Northern "Gospo zaliti nnsh chetseka kvivkv co gospo za tseyuspo."

Daris says in Teanga "Sceania uistemhens mruiund engiarruth fhaathfhei amb sceania tho iorrstunia."

Daris says in Eme "Habdin muspilmina zilgam sihirnad duhgidleh ha habdin nita hirkusdun."

Daris says in Cat "Nnmmr eoweowee eoweow ahhssser ewngew hss nnmmr waka waka ssseowmmr."

Daris says in Toad, "Hrnnnkurp nrrdeepurp deepdeep croakurpree reehrnnnkree reek hrnnnkurp nrr urpnrrurp."

:p

-- Christopher Allen

[Kristine]

Toad? There's a toad language?! :eek: :confused:

I've only seen a toad once in the castle, and that was because I was looming over my mum's shoulder watching her play. :p

[Angron]

:)

Point taken, Chris, but notice how many more repetitions in the sounds occur in the languages such as Western, cat, toad, and Northern than in Teanga, Eme or Eastern.

It is particularly noticeable in Western, and in Northern it is the reason why the word 'Charmiam' and the word 'beautiful' both translate into the word 'Ev'. In Teanga, this supposedly occurs to a lesser extent, although both the word 'he' and 'is' translate into the same word.

That's what Michael told me regarding the languages, from whom virtually all of my knowledge regarding the languages comes from.

-Zack

[Dariel]

*listens*

*squints*

*mutters* "He speaks Eme... in public..."

*asks courier for scroll*


Have a nice day.

[Michael]

Toad was mostly made for players-who-get-turned-to-toads (it can happen) and other magical stuff.

Unfortunately, when the languages were first 'put together', the directional languages (Eastern, Northern, and particularly Western) had relatively few phoneme-ish bits. I think we didn't quite expect they would be widely used; certainly, they were originally just 'test' languages. Northern has 40 phoneme-ish components; cat has 38 (including 'waka waka'), Eastern has 22; Western has 10; toad has 6. In comparison, Teanga and Eme each have over 300 phoneme-ish bits.

There must be a lot of body language and unspecified (in the written output) tonal inflection to the Western language. Certainly, all those 'm's in a row can't be too useful otherwise.

We could add more phoneme/syllable chunks to the Western list, but then the language would translate differently ... Important note to future language developers: try to have at least 100 word chunks in your list.

[Stix]

Ahhh, Jester Dagonet must know Toad, line up about 20 toads and riddle them to death! :eek: :D ;)

[Nom]

I'm not convinced that changing the mappings at this point would be such a big loss. While we LPA types make a big deal out of "learning words" IC, the fact is that we deliberately overlook quite a few curious "features" of the language system.

- Mappings are degenerate ("tonal or inflection differences"), so you can't reliably back-translate the gibberish anyway.

- Unblessed speech is not translatable. If I copy in plain-speech what someone else has said in a language, even a fluent speaker will understand none of it.

- Only inexperienced listeners, and not the speaker, ever hear unmapped tongue! (It's good to have a total newbie in your language class to OOC repeat back your words).

- Poor speakers, when speaking, produce the correct mapping, but it seems to not be translated.

- Sometimes word forms run against common sense. In Teanga, "No" is a four syllable word, whereas in reality these words are normally quite short.

- You can "translate" the common alphabet into foreign languages, and it translates to WORDS, not characters. In fact, this was my first clue to how the language algorithm really works.


My summary comments would be:

(1) If the mappings are going to be changed, make sure you tell the LPA people, and try to do it once only. However, I don't think a change is really needed.

(2) Three language "fixes":
- Speakers should be able to hear the translated form of what they say.
- It should be possible (though perhaps not mandatory) for an experienced listen to get both mapped and unmapped versions.
- Poor speakers should have a scrambler added to their speech, so the output mapping is actually wrong.


Perhaps:

Daris says in Eastern "Or'un iswainel nulnul eridoningol winianer ad or'un is oninis'un."
(The quick brown fox jumped over the lazy dog)

or is that too real-estate intensive?